Proteomic Analysis of Laser-Captured Paraffin ... · A Molecular Portrait of Head and Neck Cancer Progression ... Supplementary data for this article are available at Clinical Cancer
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Proteomic Analysis of Laser-Captured Paraffin-EmbeddedTissues:AMolecular Portrait of Head and Neck Cancer ProgressionVyomesh Patel,1Brian L. Hood,2 Alfredo A.Molinolo,1Norman H. Lee,4 Thomas P. Conrads,2
John C. Braisted,5 David B. Krizman,3 Timothy D. Veenstra,2 andJ. Silvio Gutkind1
Abstract Purpose: Squamous cell carcinoma of the head and neck (HNSCC), the sixth most prevalentcancer amongmenworldwide, is associated with poor prognosis, whichhas improved onlymar-ginally over the past three decades. A proteomic analysis of HNSCC lesions may help identifynovel molecular targets for the early detection, prevention, and treatment of HNSCC.Experimental Design: Laser capture microdissection was combined with recently developedtechniques for protein extraction from formalin-fixed paraffin-embedded (FFPE) tissues and anovel proteomics platform. Approximately 20,000 cells procured from FFPE tissue sections ofnormal oral epithelium and well, moderately, and poorly differentiated HNSCC were processedfor mass spectrometry and bioinformatic analysis.Results: A large number of proteins expressed in normal oral epithelium and HNSCC, includingcytokeratins, intermediate filaments, differentiation markers, and proteins involved in stem cellmaintenance, signal transduction, migration, cell cycle regulation, growth and angiogenesis, ma-trix degradation, and proteins with tumor suppressive and oncogenic potential, were readilydetected. Of interest, the relative expression of many of these molecules followed a distinct pat-tern in normal squamous epithelia and well, moderately, and poorly differentiated HNSCC tumortissues. Representative proteins were further validated using immunohistochemical studies inHNSCC tissue sections and tissue microarrays.Conclusions:The ability to combine laser capture microdissection and in-depth proteomic anal-ysis of FFPE tissues provided a wealth of information regarding the nature of the proteinsexpressed in normal squamous epithelium and during HNSCC progression, which may allow thedevelopment of novel biomarkers of diagnostic and prognostic value and the identification ofnovel targets for therapeutic intervention in HNSCC.
Head and neck squamous cell carcinoma (HNSCC) is the sixthmost frequent cancer in the United States and the fourth mostprevalent cancer among men worldwide (1). The prognosis ofHNSCC patients is relatively poor, largely due to the advancednature of the disease at the time of diagnosis. The identificationof the molecular mechanisms underlying HNSCC initiationand progression could aid in the development of newdiagnostic and treatment options for this disease (2, 3). In thisregard, analysis of mRNA transcripts using high-throughputgene array analysis has helped identify numerous moleculesthat may contribute to cancer development (reviewed in ref. 4).A shortcoming of this approach, however, stems from thediscordance between transcript levels and protein abundance ina highly complex and readily changing disease microenviron-ment such as cancer (5). A proteomic analysis is urgentlyneeded, as it allows the comprehensive assessment of thedistinct molecular profile of each cancer type, thus affordingthe opportunity of identifying novel prognostic markers andtherapeutic targets (6).Improved protein extraction protocols (7) combined with
recently developedmass spectrometry (MS) techniques and fullyannotated genomic databases has allowed the identificationof trace amounts of proteins present in complex samples (8).
Human Cancer Biology
Authors’Affiliations: 1Oral and Pharyngeal Cancer Branch, National Institute ofCraniofacial and Dental Research, NIH, Bethesda, Maryland; 2Laboratory ofProteomics and Analytical Technologies, Science Applications InternationalCorporation-Frederick, Inc., National Cancer Institute, Frederick, Maryland;3Expression Pathology Incorporated, Gaithersburg, Maryland; 4Department ofPharmacology and Physiology,The GeorgeWashington University Medical Center,Washington, District of Columbia; and 5Pathogen Functional Genomics ResourceCenter, J. CraigVenter Institute, Rockville, MarylandReceived 6/8/07; revised10/19/07; accepted12/5/07.Grant support: Intramural Program, National Institute of Dental and CraniofacialResearch and National Cancer Institute, NIH, under contract NO1-CO-12400.The costs of publication of this article were defrayed in part by the payment of pagecharges.This article must therefore be hereby marked advertisement in accordancewith18 U.S.C. Section1734 solely to indicate this fact.Note: Supplementary data for this article are available at Clinical Cancer ResearchOnline (http://clincancerres.aacrjournals.org/).The content of this publication does not necessarily reflect the views or policies ofthe Department of Health and Human Services, nor does mention of trade names,commercial products, or organization imply endorsement by the U.S. Government.Current address for B.L. Hood and T.P. Conrads: Clinical Proteomics Facility,University of Pittsburgh Cancer Institute, Pittsburgh, Pennsylvania.Requests for reprints: J. Silvio Gutkind, Oral and Pharyngeal Cancer Branch,National Institute of Dental and Craniofacial Research, NIH, 30 Convent Drive,Building 30, Room 211, Bethesda, MD 20892-4330. Phone: 301-496-6259;Fax: 301-402-0823; E-mail: [email protected].
F2008 American Association for Cancer Research.doi:10.1158/1078-0432.CCR-07-1497
www.aacrjournals.orgClin Cancer Res 2008;14(4) February15, 2008 1002
Furthermore, the combination of proteomic analysis with lasercapture microdissection may afford performing the proteomiccharacterization of normal and pathologic cell populations fromclinical specimens, thus reflecting their protein make up as theyexist in vivo (9). We have recently conducted the first proteome-wide analysis of microdissected frozen HNSCC tissue (10).Although protein recovery was acceptable, challenges related tothe poor histology and limited availability of appropriatelypreserved frozen tissue samples still remain.Formalin fixation and tissue embedding in paraffin wax
(FFPE) is a universal approach for tissue processing, histologicevaluation, and routine diagnosis, as it preserves the cellularmorphology and tissue architecture. FFPE clinical specimens,however, are not routinely used for MS-based proteomic studiesbecause formaldehyde-induced cross-linking renders proteinsrelatively insoluble and unsuitable for extraction and subse-quent MS analysis (11). In this study, we combined the use oflaser capture microdissection, newly developed proceduresenabling the extraction of peptides directly from FFPE samples,and optimized chromatographic approaches to undertake alarge-scale proteomic study to identify proteins expressed inFFPE HNSCC tissues. We show that this novel proteomicsplatform enables the identification of hundreds of proteinsexpressed in normal oral epithelium and cancerous HNSCClesions. A large number of differentiation markers, stem cellproteins, and molecules that are likely to play key roles inaberrant cell growth, including proteins involved in cell cyclecontrol, angiogenesis, and metastasis, were identified. Theseproteins may represent novel biomarkers with diagnostic andprognostic value for HNSCC, as well as new potential moleculartargets for pharmacologic intervention in this disease.
Materials andMethods
Tissues. Appropriate paraffin blocks of formalin-fixed tissues fromnormal squamous epithelium (n = 4), and well differentiated (WD;n = 4), moderately differentiated (MD; n = 4), and poorly differentiated
(PD; n = 4) HNSCC were retrieved from the Head and Neck CancerTissue Array Repository (TMA)6 (12), under an approved ResearchActivity Involving Human Subjects protocol, from the Office of HumanSubjects Research, NIH, involving the use of anonymous normal andHNSCC tissues. Before analysis, H&E-stained section from each samplewas evaluated and the suitability of inclusion for the study wasdetermined. Five-micrometer sections were used for all subsequentanalysis.
Immunohistochemistry. Primary antibodies used for validationstudies include mouse anti–desmoglein-3 (Invitrogen; 1:25), rabbitanti-cytokeratin 4 (Abcam; 1:100), mouse anti-cytokeratin 16 (Abcam;1:25), rabbit anti-desmoplakin antibody (Abcam; 1:200), and mouseanti-vimentin (Dako; 1:100). Secondary antibodies conjugated withbiotin (Vector) were used, diluted to 1:400. Tissue slides containingarchival FFPE sections, or tissue microarrays (TMA)6 consisting of 508HNSCC and controls, were processed for immunohistochemicalanalysis as described in Supplementary Material.
Laser capture microdissection and protein extraction. FFPE oralcancer tissue sections were deparaffinized in SafeClear II, hydrated,and stained in Mayer’s hematoxylin followed by dehydration asdescribed in Supplementary Material. For laser capture microdissection,stained uncovered slides were air dried and f20,000 cells werecaptured onto CapSure LCM Caps (MDS Analytical Technologies) usinga PIXCELL IIe microdissection equipment (MDS Analytical Technolo-gies). Caps were transferred to a 0.5 mL sterile Eppendorf tube forprotein extraction using the Liquid Tissue MS Protein Prep kit accordingto the manufacturer’s protocol (Expression Pathology, Inc.), asdescribed in Supplementary Material.
Tandem MS and bioinformatic analysis. FFPE-extracted samples wereprocessed, quantified, and used for nanoflow reversed-phase liquidchromatography followed by tandem MS, as described in Supplemen-tary Material. The spectra were searched against the UniProt humanprotein database (06/2005 release) from the European BioinformaticsInstitute7 using SEQUEST (Thermo Electron). Results were furtherfiltered using software developed in-house to determine uniquepeptides and proteins, which has a predicted error <1.5% (13).
Table 1. Clinical features of HNSCC cases and summary of the liquid chromatography–tandem MS data
Sample Pathology Sex Age Location Total peptides Unique proteins Total (n = 4)
2N N M 71 Oral 1,143 2078N N U 50 Oral 775 14036N N M 63 Oral 620 13737N N M 55 Oral 339 106 39112A WD SCC F 62 Oral 1,454 37614A WD SCC F U Oral 1,570 35615A WD SCC F 60 Oral 1,586 35116A WD SCC F 42 Oral 1,721 323 8661A MD SCC F 67 Oral 1,833 3662A MD SCC M 56 Oral 1,422 3226A MD SCC M U Oral 838 2169A MD SCC M 33 Oral 795 194 72910A PD SCC F 52 Oral 1,202 23911A PD SCC M 72 Oral 1,411 29613A PD SCC U 50 Oral 1,073 25718A PD SCC U U Oral 774 185 676
NOTE: All clinical samples were retrieved from the Head and Neck Tissue Microarray initiative, and chosen based primarily on location within theoral cavity. Samples were assessed for the presence of normal and malignant squamous epithelia that were either WD, MD, or PD. Whereavailable, information on patient gender and age is included. Summary of number of peptides and proteins detected in each sample is included.Abbreviations: N, normal; M, male; F, female; U, unknown.
To facilitate the biological interpretation of the extensive protein listsgenerated in these studies, the protein accession numbers were used toclassify the proteins in Gene Ontology categories, based on theirbiological process and molecular and cellular functions (14), and toperform Expression Analysis Systematic Explorer analysis,8 whichenables the discovery of enriched biological themes within gene/protein lists, and the generation of protein annotation tables.
Results
HNSCC samples and laser capture microdissection. To gaininsight into the nature of proteins expressed during HNSCCprogression, we conducted a proteomic analysis of FFPEHNSCCtumors arising within the oral cavity. Tissue samples wereclassified using light microscopic examination of H&E-stainedsections into normal (i.e., oral stratified epithelia lacking
malignant features) or squamous carcinomas that were WD,MD, or PD. Each group consisted of four independent samples.The available clinical information is included in Table 1. Theanalytic workflow for the overall study is depicted in Fig. 1,which is the result of combining two different technologyplatforms, laser capture microdissection, and tandem MS. Asshown schematically (Fig. 1A) and in detail (Fig. 1B), lasercapture microdissection is well suited for the rapid procure-ment of specific cell populations, which are captured onto capsfor immediate processing and analysis. The complexity ofthe resulting peptide mixture extracted from each sample isexemplified in Fig. 1A by the base peak spectrum of a repre-sentative WD HNSCC tumor (right). A tandem MS spectrumidentifying a peptide originating from vimentin is also shown inthis figure. This proteomic platform results in broad dynamicrange of peptide measurements, which may aid in theidentification of important molecules involved in squamouscarcinogenesis as well as biomarkers for HNSCC progression.8 http://david.abcc.ncifcrf.gov
Fig. 1. Work flow for protein analysis ofFFPE oral squamous cell carcinoma.A, different steps involved in the process ofprotein analysis from laser-captured FFPEcancer tissues, including (from left to right)tissue biopsy, laser capture microdissection,sample preparation, and analysis byreversed-phase liquid chromatography ^tandemMS.The latter involves the initialseparation of complex peptide mixturesby Nanoflow reversed-phase liquidchromatography, followed by LITMS. Acomplex base peak chromatogram of arepresentative HNSCC case is included, aswell as the tandemMS spectra of a selectedpeptide whose identity was confirmed asvimentin.B, laser capturemicrodissectionofaWDFFPE oral squamous cell carcinoma(top left).The area of interest is pulsed withlaser (bottom left) and captured cells wereretrieved on a cap (bottom right). Remnanttissue remains on the slide (top right).
Human Cancer Biology
www.aacrjournals.orgClin Cancer Res 2008;14(4) February15, 2008 1004
Protein abundance in normal and tumor HNSCC. Samplescontaining f20,000 cells were processed as described inMaterials and Methods and analyzed using gas phase fraction-ation in the m/z dimension (GPFm/z). To enable the compar-ison of protein expression across all tissue samples, we used aspectral count method in which each protein that wasidentified in any one set was evaluated based on the numberof unique peptides identified, and the total number of timesthat those peptides were observed in five successive runs foreach set (Table 1). The number of proteins identified for eachset as a group (all four sample results pooled together) rangedfrom 391 in the normal oral epithelia to 866 total proteins inthe WD HNSCC tumors.The data sets for each group of clinical samples were sorted
based on the number of peptides identified in all tissuescombined and the percentage of these total peptides detected innormal and each of the classified HNSCC tissues. Thoseproteins detected in normal and all HNSCC tissues are listedin Table 2 and sorted based on the number of identifiedpeptides. As indicated, 115 proteins were identified as sharedacross all normal and tumor tissues. The utility of this overallapproach is reflected in the identification of 49 peptides forglyceraldehyde-3-phosphate dehydrogenase, a protein productof a housekeeping gene routinely used to normalize geneexpression experiments. In spite of observing fewer overallpeptides in normal oral epithelia, which may have resulted froma reduced protein recovery from these samples, the relativedetection level of glyceraldehyde-3-phosphate dehydrogenasewas nearly equal across all tissue groups analyzed in this study.Thus, it is also possible that normal tissues might exhibit agreater diversity of molecules, therefore fewer achieving thelowest limit of detection of our current analytic method.Two groups of proteins, cytokeratins and desmosomal
proteins, stood out as the most abundant in all four-tissuesets. With few notable exceptions, most keratins were nearlyequally represented across all tissue samples. For example,cytokeratin 5, a keratin expressed in basal layer of normalstratified epithelia, was the most abundant and was identifiedby a similar number of peptides in each tissue group.Cytokeratin 14, another basal keratin, was less abundant innormal tissues. Conversely, cytokeratin 4, which is expressed inthe nonkeratinazing layer of stratified epithelium, was moreabundant in the normal samples. Desmosomal proteins,including desmoplakin, plakophilin 1, periplakin, and desmo-glein precursors, were prominent in all samples. Also notable inthis list were the family of calcium-binding proteins (calgranu-lin A, calgranulin B, S100 A14, calmodulin-like protein),keratinocyte differentiation markers (involucrin, small pro-line-rich protein 3, profilin 1, cornifin A, cornifin B), and manymembrane-related molecules (Annexin A1, actin-like protein2). A number of proteins including heat shock protein 27(HSP27) (15), HSP70 (16), and glutathione S-transferase (15),expected to be of higher abundance in tumor samples, wereidentified. A number of signaling molecules (e.g., Ras GTPase-activating-like protein IQGAP1, obscurin, tyrosine-proteinkinase ITK) were also identified. Conversely, the DNA excisionrepair protein ERCC-5 was found to be less abundant in tumorsamples. These data indicate that the analysis of FFPE tissuesenables the identification of a broad range of functionallydiverse proteins in normal and tumor squamous epithelia,albeit many exhibit a distinct expression profile.
Proteins detected only in HNSCC. Having identified com-mon proteins in both normal and tumor samples, a list ofmolecules only identified in tumor samples was collated(Table 3). Forty-two proteins unique to HNSCC were readilydetected by multiple peptides (>10). Among them, the mostabundant was vimentin, a protein involved in epithelial-mesenchymal transition. A variety of proteins involved in cellmigration, signaling, and proteolysis were also identified.Eighty-five less abundant proteins (i.e., identified by >4peptides but <10 peptides) across all tumor samples werealso detected. This group included proteins involved in DNAsynthesis, metabolism, and cell signaling. These data providea list of proteins that may play a putative role in tumorprogression.
Proteins detected in normal oral squamous epithelium. Togain insight into the proteins identified exclusively in normalsamples, the initial data set was sorted to filter out moleculesnot found within the tumors, in descending order based onpeptide numbers (Supplementary Table S1). The proteins inthis list were identified by fewer peptides, reflecting their lowerabundance in normal oral squamous epithelium. Interestingly,proteins identified by z3 unique peptides included the low-density lipoprotein receptor-related protein 12 precursor, alsoknown as suppressor of tumorigenicity protein 7, and twoparticular proteins identified with a single peptide included,activin h B chain and adenomatous polyposis coli protein. Thedata indicate that a subset of proteins are expressed preferen-tially in normal tissues, indicating that they could play a role inany of the biological functions done by normal stratified oralepithelium, including maintenance of normal differentiationprogram and tumor suppression.
Proteins detected in differentiated tumor tissues. Proteins ofinterest detected exclusively in the WD tumors, albeit with fewpeptides, included those involved in the dynamic function of thecytoskeleton, as well as molecules stimulating the Notchpathway, such as Delta 4 and Delta 1 (Supplementary TableS2). Proteins of interest and detected only in the MD group withone peptide include Wilms’ tumor-associated protein and eso-phageal cancer–related gene-coding leucine-zipper motif, andunusual cadherins and desmosomal proteins for epithelial cellssuch as placental cadherin and protocadherin g A6 (Supplemen-tary Table S3). Interesting proteins detected only the PD groupinclude the potential oncoprotein AF1q (17), two peptidesderived from epithelial protein lost in neoplasm, and numerousproteins involved in cell cycle control and fatty acid metabolismand membrane trafficking (Supplementary Table S4).
Comparisons of proteins between samples. The list of proteinsidentified in each group was also compared to explore whetherthey could increase our understanding of HNSCC pathogenesisand its progression. Proteins identified only in normal and WDtumors by z3 peptides included the stem cell protein PIWIL3(ref. 18; Table 4; Supplementary Table S5). A number ofproteins were common to WD and MD tumors, represented byoncoprotein DJ-1 (Supplementary Table S6). Similarly, com-mon proteins between MD and PD tumors included the tumorpromigratory protein JWA. Notably, low-abundance proteinsshared by the same group of HNSCC tumors also includedsignal transducer and activator of transcription 3, its activatingkinase, Janus-activated kinase 2, and the key translational regu-lating protein mammalian target of rapamycin (Supplemen-tary Table S7), which are implicated in HNSCC progression
The Oral Cancer Proteome
www.aacrjournals.org Clin Cancer Res 2008;14(4) February15, 20081005
(19, 20). Full lists of proteins for these comparisons areavailable in Supplementary Tables S5 to S7.Gene ontology analysis. Analysis of each sample set resulted
in the identification of a large number of proteins. Therefore, itwas necessary to integrate these data based on our currentlyavailable knowledge of biological functions to determine theirindividual biological roles and to recognize categories ofproteins that may be underrepresented or overrepresented intumors. For this, we used bioinformatic tools to classify
proteins across each of the HNSCC tumor types into differentgene ontology categories. This approach enabled the examina-tion of the overall cellular compartment in which the identifiedproteins are expected to reside and function. Although aminimal difference was observed between normal and HNSCCsamples, we found that 133-295, 10-21, 77-185, and 86-184proteins were associated with the cytoplasm, extracellularmatrix, membrane, and the nucleus, respectively. Analysis ofthe biological processes in which these proteins act revealed
Table 2. Proteins identified in both normal and tumor HNSCC tissues (Cont’d)
Accession Protein Peptides(number)
Normal(%)
WD(%)
MD(%)
PD(%)
P62913 60S ribosomal protein L11 (CLL-associated antigen KW-12) 17 5.88 47.06 23.53 23.53P62899 60S ribosomal protein L31 16 6.25 43.75 18.75 31.25Q99456 Keratin, type I cytoskeletal 12 (cytokeratin 12; K12; CK 12) 14 7.14 57.14 21.43 14.29Q02413 Desmoglein 1 precursor (desmosomal glycoprotein 1; DG1; DGI) 14 14.29 35.71 42.86 7.14P83731 60S ribosomal protein L24 (ribosomal protein L30) 13 7.69 38.46 23.08 30.77P46940 Ras GTPase-activating-like protein IQGAP1 (p195) 13 7.69 30.77 30.77 30.77P00505 Aspartate aminotransferase, mitochondrial precursor (transaminase A) 13 15.38 23.08 7.69 53.85P78371 T-complex protein 1, h subunit (TCP-1-h; CCT-h) 13 15.38 30.77 30.77 23.08P50990 T-complex protein 1, u subunit (TCP-1-u; CCT-u) 13 23.08 38.46 15.38 23.08O60506 Heterogeneous nuclear ribonucleoprotein Q (hnRNP Q) 12 8.33 41.67 41.67 8.33Q00610 Clathrin heavy chain 1 (CLH-17) 12 16.67 33.33 8.33 41.67O60437 Periplakin (195 kDa cornified envelope precursor protein) 12 25.00 25.00 25.00 25.00Q07020 60S ribosomal protein L18 12 33.33 33.33 25.00 8.33P30041 Peroxiredoxin 6 (antioxidant protein 2) 11 18.18 18.18 27.27 36.36Q96AA2 Obscurin 11 18.18 36.36 27.27 18.18Q5U4P6 KHSRP protein 11 18.18 45.45 9.09 27.27Q08881 Tyrosine-protein kinase ITK/TSK 11 18.18 36.36 27.27 18.18P13639 Elongation factor 2 (EF-2) 11 27.27 45.45 9.09 18.18P35321 Cornifin A (small proline-rich protein IA; SPR-IA; SPRK) 11 27.27 45.45 9.09 18.18P22528 Cornifin B (small proline-rich protein IB; SPR-IB) 11 36.36 36.36 9.09 18.18Q01518 Adenylyl cyclase-associated protein 1 (CAP 1) 10 10.00 30.00 20.00 40.00P08572 Collagen a 2(IV) chain precursor 10 10.00 50.00 20.00 20.00Q96MG1 Hypothetical protein FLJ32421 10 10.00 50.00 30.00 10.00O75312 Zinc-finger protein ZPR1 (zinc finger protein 259) 10 30.00 10.00 10.00 50.00Q8TBA0 Chromosome 8 open reading frame 21 10 30.00 20.00 10.00 40.00Q96CN5 Hypothetical protein MGC20806 10 30.00 40.00 10.00 20.00P40121 Macrophage capping protein (actin-regulatory protein CAP-G) 9 11.11 11.11 44.44 33.33P06733 a Enolase (Enolase 1) 9 11.11 22.22 33.33 33.33P12035 Keratin, type II cytoskeletal 3 (cytokeratin 3; K3) 9 11.11 44.44 22.22 22.22O00299 Chloride intracellular channel protein 1 (nuclear chloride ion channel 27) 9 11.11 44.44 22.22 22.22Q02388 Collagen a 1(VII) chain precursor (long-chain collagen; LC collagen) 9 22.22 11.11 11.11 55.56Q9Y6X9 Zinc finger CW-type coiled-coil domain protein 1 9 22.22 55.56 11.11 11.11P04080 Cystatin B (liver thiol proteinase inhibitor; CPI-B; Stefin B) 9 33.33 33.33 22.22 11.11P49368 T-complex protein 1, g subunit (TCP-1-g; CCT-g) 8 12.50 12.50 25.00 50.00Q6ZT17 Hypothetical protein FLJ45043 8 12.50 25.00 25.00 37.50O15353 Forkhead box protein N1 (transcription factor winged-helix nude) 8 37.50 25.00 25.00 12.50P28715 DNA excision repair protein ERCC-5 8 50.00 25.00 12.50 12.50P24534 Elongation factor 1-h (EF-1-h) 7 14.29 14.29 28.57 42.86Q8N7I6 Hypothetical protein FLJ25506 6 16.67 16.67 33.33 33.33P62277 40S ribosomal protein S13 6 16.67 16.67 16.67 50.00Q99551 Transcription termination factor, mitochondrial precursor (mTERF) 6 16.67 33.33 33.33 16.67P31153 S-adenosylmethionine synthetase g form (AdoMet synthetase) 6 16.67 50.00 16.67 16.67Q9UKG9 Peroxisomal carnitine O-octanoyltransferase (COT) 6 33.33 16.67 33.33 16.67Q15369 Transcription elongation factor B polypeptide 1 (elongin C) 6 50.00 16.67 16.67 16.67Q06830 Peroxiredoxin 1 (EC 1.11.1.-; thioredoxin peroxidase 2) 5 20.00 20.00 40.00 20.00O75151 PHD finger protein 2 (GRC5) 4 25.00 25.00 25.00 25.00O60231 Putative pre-mRNA splicing factor RNA helicase (DEAH-box protein 16) 4 25.00 25.00 25.00 25.00P52907 F-actin capping protein a-1 subunit (CapZ a-1) 4 25.00 25.00 25.00 25.00Q99698 Lysosomal trafficking regulator (Beige homologue) 4 25.00 25.00 25.00 25.00Q8TBU6 Hypothetical protein FLJ11848 4 25.00 25.00 25.00 25.00
NOTE: Proteins that were identified as common in all normal and HNSCC tissues were sorted across data sets of the different tissue samples asdescribed in Materials and Methods, based on the corresponding peptide number in descending order and their relative distribution (%) acrossthe different samples: normal, WD, MD, and PD HNSCC. Accession number for each protein is also indicated.
The Oral Cancer Proteome
www.aacrjournals.org Clin Cancer Res 2008;14(4) February15, 20081007
several groups of molecules that were highly represented intumors, many involved in cell adhesion, cell cycle, celldifferentiation, cellular metabolic process, cell motility, endo-peptidase activity, signal transduction, and gene transcriptionand translation, closely aligned with the processes of carcino-genesis. Indeed, a large number of proteins belonging to thesefunctional groups were ascribed to tumors, and to their normaltissue counterpart, albeit to a lesser extent (SupplementaryTable S8). A detailed list of molecules predicted to participate insignal transduction (238 proteins), protein phosphorylation(39 proteins), endopeptidase activity (35 proteins), cellmotility (47 proteins), cell cycle regulation (99 proteins), andcell adhesion (71 proteins) across normal tissues and thedifferent tumor phenotypes is provided in SupplementaryTables S9 to S14.Detection of novel proteins in normal and tumor HNSCC. A
key advantage afforded by this proteome-wide analysis is thepotential for discovery of new molecules yet to be described fornormal and malignant oral squamous tissues. Several hypo-
thetical novel proteins, whose existence is predicated oncomputational analysis of open reading frames (21), werefound to be present in normal and tumor samples. Theseproteins were collated and organized in descending order basedon peptide number. As seen in Supplementary Table S15, themajority of the proteins with z4 peptides were detected ineither tumor or normal samples. Collectively, the emerginginformation identified many novel proteins in normal andtumor samples, whose nature can now be analyzed. In thisregard, structural features of these previously predicted proteinsare quite diverse and suggestive of multiple roles in signaltransduction, cell communication, and secretion, amongothers. Future work could help establish their possible role inHNSCC development and progression.
Proteins identified as of interest for HNSCC. The ability todiscriminate subsets of proteins differentially abundant withinnormal oral squamous epithelia and tumors exhibiting distinctdifferentiation characteristics provided an opportunity to minethe data to identify proteins of interest as well as putative
Table 3. Proteins identified only in tumoral HNSCC tissues (Cont’d)
NOTE: Proteins identified exclusively in the tumor samples were sorted as described in Materials and Methods based on their peptide number indescending order and their relative distribution (%) across the different samples: WD, MD, and PD.
The Oral Cancer Proteome
www.aacrjournals.org Clin Cancer Res 2008;14(4) February15, 20081009
biomarkers of HNSCC development and progression. Aselected group of molecules identified as being differentiallyabundant were sorted by total peptide number and theirrelative distribution across the normal and tumor samples(Table 4). Notably, the peptide distribution of the cytokeratins14, 17, and 16 was observed to be lower in the normal samplescompared with the WD, MD, and PD tumor samples. Bycontrast, the peptide distribution of cytokeratin 4 was higher innormal (f77%) compared with the tumor samples (f4-12%).Also included in the list are cytokeratins 7 and 18 albeit with alower peptide number. The relative distribution indicates thatthese molecules are less abundant in normal samples than inMD and PD tumors, respectively. Molecules involved in cell-to-cell interaction were also identified as potential biomarkers andinclude desmoplakin, democollin 2A/2B, demoglein 3 precur-sor, plakophilin 1, and plakophilin 3. The total peptide numberfor this group of proteins ranged from 7 to 385, with a relative
similar distribution in all samples with the exception ofdemocollin 2A/2B, which was undetectable in normal samples.Furthermore, from the total peptide number of 385 fordesmoplakin, f44% of these were detected in well-differenti-ated tumors and in the other samples indicating close to equallevels. Other proteins of interest include HSP27 and HSP70,vimentin, glutathione S-transferase, and integrin h4. In all theseproteins, the distribution of peptides was low to undetectable innormal samples when compared with the tumors. Among thelatter, we found proteins that may play a role in tumor progres-sion, such as SET protein (phosphatase 2A inhibitor I2PP2A)and ELAV-like protein 1 (Hu-antigen R), and many surfaceproteins, including 4F2 cell surface antigen heavy chain (CD98),and, as described above, integrin h4 (CD104 antigen), whichmay represent promising markers to study tumor progression.Validation of biomarkers by immunohistochemistry. Having
identified proteins of potential interest to HNSCC progression,
Table 4. Representative proteins of interest for HNSCC
Accession Protein Peptides (number) Normal (%) WD (%) MD (%) PD (%)
NOTE: Known proteins from the total list that may represent proteins of interest as well as potential biomarkers were sorted based on their totalpeptide number and relative distribution (%) across the normal and WD, MD, and PD tumor samples.
Human Cancer Biology
www.aacrjournals.orgClin Cancer Res 2008;14(4) February15, 2008 1010
a subset of these were chosen for validation based on therelative abundance of corresponding peptides in the differentsamples analyzed. Validation studies were conducted usingstandard immunohistochemistry in archival HNSCC tissuesfollowed by in-depth analysis of hundreds of HNSCC casesusing a recently developed HNSCC-specific TMA. As seen inFig. 2A, cytokeratin 4 was detected predominantly in thesuprabasal layers of the control normal tissues, whereas theexpression of this molecule in HNSCC tumors was restrictedto only few cells in <20% of WD tumors. Most MD to WDcells stained positive for cytokeratin 16, whereas in the normaltissues this protein was present in the suprabasal layers of theoral squamous epithelium (second from top). Desmoplakin(third from top) was found to be predominantly membranous,with a higher immunoreactivity in the suprabasal areas of thesquamous epithelium in normal tissues. Tumor cells showedintense staining of the membrane together with a moderate tostrong cytoplasmic staining, distributed mostly along the moredifferentiated areas of all the tumor cases examined. Analysisof desmoglein-3 (fourth from top) showed sharp stainingwithin the membrane in normal oral squamous mucosa witha stronger signal in suprabasal layers. Desmoglein-3 was alsopositive in all tumor samples, following a membranedistribution and a less intense cytoplasmic signal. Immuno-detection of vimentin (bottom) was positive in only a fewnormal isolated cells with very distinct dendritic-like mor-phology in the oral squamous epithelium, whereas theunderlying stromal cells were all intensely stained. By contrast,a high proportion of the malignant squamous cells in thetumor cases showed increased immunoreactivity to vimentin.Collectively, the data indicate that the profile of proteinexpression identified using MS is reflected in an independentarchival HNSCC sample set.The previous data indicated that a subset of proteins
(cytokeratin 4, cytokeratin 16, vimentin, and desmoplakin)could effectively distinguish the differing differentiation com-partments in archival HNSCC tissues. Therefore, we chose touse an oral cancer–specific TMA for the high-throughputstaining and scoring of these predictive immunohistochemicalmarkers in this cancer type. As shown in Fig. 2B (top), althoughcytokeratin 4 was poorly expressed in the majority of thetumors (left and inset), strong staining for cytokeratin 16 inHNSCC was observed almost exclusively in WD tumors orotherwise WD areas (second from left and inset), as most MDand PD tumors failed to react strongly. Vimentin immunore-activity was almost exclusively limited to a subset of malignanttumors, and in the majority of the cases this staining was focal(third from left and inset). Finally, staining for desmoplakin wasstrongly positive in normal tissues as well as in almost all tumorsamples (right and inset).A semiquantitative analysis was applied to the TMA staining.
As indicated in Fig. 2B (bottom), cytokeratin 4 and cytokeratin16 scored positive in normal tissues based on the stainingpattern of the suprabasal layer, and their expression in tumorswas often restricted to WD areas. For the evaluation of vimentin(third from left) and desmoplakin (right), we classified cellstaining in each tissue core as positive and negative because noclear correlation with tumor differentiation was noted from theinitial analysis. Of note, vimentin and desmoplakin wereassessed to be negative and strongly positive, respectively, in allthe relevant normal oral epithelial tissues analyzed.
Discussion
In this study, we describe the utility of a novel proteomicsplatform for the global detection of expressed proteins in FFPEtissues and its use for biomarker discovery and identification ofproteins that may contribute to HNSCC development andprogression. This approach enabled identification of a largenumber of molecules, including cytokeratins and intermediatefilament proteins, differentiation markers, proteins involvedin stem cell maintenance, signal transduction and cell cycleregulation, growth and angiogenic factors, matrix-degradingproteases, and proteins with tumor suppressive and oncogenicpotential. Of interest, detection and relative expression of manyof these molecules followed a distinct pattern in normalsquamous epithelia and WD, MD, and PD HNSCC tumortissues. The ability to correlate protein expression profiles withhistopathologic classification of disease may allow the devel-opment of novel biomarkers of diagnostic and prognostic valueand may help identify novel targets for therapeutic interventionin HNSCC.
Certain advantages embedded within the workflow devel-oped for this study include the efficient solubilization anddigestion of proteins from FFPE archival tissue withoutfractionation, such that they are amenable for identificationusing tandem MS for a complete proteomic representation (7).Furthermore, optimization of combining laser capture micro-dissection with shotgun proteomic technologies enabled thedetection of proteins expressed primarily within the tumor cellsrather than in the stroma and other complex cellularcomponents of the tumor microenvironment. Although theexactness of MS identifications can be challenging, particularlyfor low abundant proteins, current MS instrumentation andimproved bioinformatic capabilities provide a high probabilityof protein identification (22). Thus, rather than reportingproteins identified by two or more unique peptide sequencematches, a common practice in the field that may sacrificemany correct protein identifications derived from high-qualitysingle hits, we chose to report here all proteins identified basedon stringent biophysical and statistical criteria. Indeed, theability to take advantage of new technological developmentsmay now enable the discovery of molecules that althoughpresent in low abundance may nevertheless play importantbiological roles in tumor development.
In this regard, it was encouraging that certain proteinsindicative of ‘‘proof of principle’’ could be readily detected. Forexample, the total peptide number for glyceraldehyde-3-phosphate dehydrogenase, a protein frequently used fornormalization, was nearly equally distributed across thedifferent samples, suggesting that the protein recovery wassimilar for each group of tissues. Further validation of ouranalysis was illustrated by the identification of members of thetaste receptors (T2R13, T2R38), olfactory receptors (5M11,13C4, JCG2), and oral facial proteins (oral-facial-digitalsyndrome 1 protein), each with a single peptide. Althoughthese proteins may not be involved in cancer, they arenevertheless known to be expressed in oral squamous tissues(23–25). The abundance of the cytokeratins (1, 4, 5, 7, 14,16–18) and the desmosomal proteins (desmoplakin, desmo-glein 3, desmocollin, epiplakin, plakophilins) was striking,particularly when considering the limited amount of sampleavailable for the proteomic analysis. Noteworthy, desmoglein 1
The Oral Cancer Proteome
www.aacrjournals.org Clin Cancer Res 2008;14(4) February15, 20081011
and desmoglein 3 are both expressed in the skin, but the130 kDa molecule desmoglein 3 is preferentially expressed inoral epithelium (26). As these molecules function together tomaintain structural integrity of the normal oral epithelium,changes in their relative levels might represent putativebiomarkers of disease progression (26).From these long lists of proteins identified in normal and
cancer cells, can we find those that may contribute to cancerdevelopment? The answer is likely yes, when considering thenature of many of the proteins identified that are predictedto be involved in cell cycle regulation, signal transduction,and proteolysis. For example, checkpoint kinase 1, a serine/threonine protein kinase that is a key mediator in the DNAdamage– induced checkpoint network (27), was highlydetected in normal epithelia but poorly detected in HNSCCtumors. Similarly, Apc, the protein product of the adenomatouspolyposis coli (apc) gene that prevents aberrant activity of theWnt/h-catenin signaling system and is the most frequentlymutated molecule in colon cancer (28), was detected as a singlepeptide only in normal oral epithelium. Low-density lipopro-tein receptor–related protein 12, also known as suppressor oftumorigenicity protein 7, a tumor suppressive protein whosegene is located on human chromosome 8 q22.2-23.1, a locus ofhigh polymorphism and genetic alterations in cancer biopsiesincluding HNSCC (29), was only found in normal oralepithelium. Similarly, a single peptide for cyclin K, a proteinthat acts as a regulatory subunit of CDK9 thereby regulating thetranscription of a subset of genes (30), was detected only innormal tissues. As cyclin K is regulated by p53, its loss in tumortissues may reflect the decreased p53 activity that characterizesHNSCC (3). Aligned with this possibility, individual peptidesfor a tumor suppressor gene on 17p13.3, hypermethylatedin cancer 1, HIC-1 , and a direct target for p53 that is involvedin the inhibition of cell growth and the initiation of cell deathand senescence programs in response to DNA damage (31),were identified in normal epithelial cells but only in one case ofPD tumor sample. Thus, a DNA damage sensing molecule,checkpoint kinase 1, at least two p53 targets, cyclin K andHIC-1, as well as lipoprotein receptor–related protein 12 andApc, the latter a well-known tumor suppressor protein poorlyinvestigated in HNSCC, seem to be more prominent in normalepithelial than in tumor cells. Collectively, these results suggestthe existence of a network of tumor-restricting mechanisms thatprotect the integrity of the normal squamous epithelium whoseloss or decreased expression and function may contribute toHNSCC progression.On the other hand, several proteins involved in cell cycle
progression, particularly G2-M transition and mitosis, such asseptin 9 and centromeric protein E, were only detected intumor samples, reflecting their active state of proliferation. Anunusual cell cycle regulating protein, prohibitin, which hasbeen recently observed to play an unexpected function in theactivation of Raf/MEK/ERK pathway by Ras and in modulatingepithelial cell adhesion and migration (32), was only detectedin tumors. Another surprising finding was the detection of twopeptides derived from EVI-5 oncogene. This protein was firstidentified in experimental T-cell lymphomas by retroviralinsertion strategies and has been recently shown to act togetherwith Polo-like kinase to ensure mitotic fidelity (33). Thus, bothprohibitin and EVI-5 may represent excellent candidates to playa role in aberrant cell growth in HNSCC.
Several molecules with a role in the transduction ofproliferative signals were also identified in normal and tumorHNSCC cells. For example, epidermal growth factor receptorwas detected in tumor samples but not in normal oralepithelial tissues, reflecting the overexpression of this growthfactor tyrosine kinase receptor in HNSCC (34). We also foundexpression of one epidermal growth factor receptor ligand,neuregulin-2, in HNSCC, suggesting an increased complexityof the epidermal growth factor receptor network in squamouscarcinogenesis. Similarly, numerous signaling molecules in-volved in cell migration were detected in HNSCC cells. Theyinclude several members of the Rho family of small GTPases,and few peptides derived from two novel guanine nucleotide
Fig. 2. Validation of biomarkers by immunohistochemistry on archival HNSCCtissues. A, archival tissues consisting normal and tumor HNSCC were processedand used for immunodetection of the indicated proteins with appropriateantibodies as described in Materials and Methods.
Human Cancer Biology
www.aacrjournals.orgClin Cancer Res 2008;14(4) February15, 2008 1012
exchange factors for Rho GTPases, RhoGEF 10 and RhoGEF19, in tumor cells, all suggestive of an important role for RhoGTPases and their regulatory proteins and downstream targetsin HNSCC progression. Two cell surface receptors, integrin h4
and plexin B1, which are involved in cell motility were alsoreadily detected in tumor cells. In this regard, whereas integrinh4 contributes to keratinocyte cell migration and facilitatestumor angiogenesis (35), plexin B1, which was initiallyidentified based on its role in axons guidance, is now knownto play an important role in endothelial cell migration andtumor angiogenesis (36). We have also shown that HNSCCexpress high levels of the plexin B1 ligand, semaphorin 4D(37), suggesting the existence of an autocrine plexin B1–semaphorin 4D loop that may promote HNSCC cell migra-tion. These proteins may also promote aberrant HNSCCgrowth, as suggested by recent studies indicating that integrinh4 and plexin B1 can stimulate members of the epidermalgrowth factor receptor and Met family of growth factorreceptors (38–40).
Proteases and their inhibitors form a complex proteolyticsystem and are ultimately responsible for cancer cell invasionand metastasis. In this study, the abundance of members of theADAMTS proteases family was particularly notable. Theseproteases likely contribute to extracellular matrix degradation,cell-to-cell adhesion, cell proliferation, and migration, and theprocessing of cytokines and growth factors, all aiding tumorprogression and angiogenesis. Cathepsin D, which is alysosomal aspartic protease, was also detected in the samples.Although cathepsins are involved in bulk protein turnover, theyalso have specialized roles in processes such as growth factorturnover and antigen presentation (41, 42). Cathepsin D wasthe most detectable protease in HNSCC samples, particularly inMD tumor cells. This cathepsin is often observed to be largelyoverexpressed in breast cancer tissues and their derived celllines and its expression levels correlate with the incidence ofclinical metastasis and shorter survival times (43). Furthermore,cathepsin D overexpression increases the growth and metastaticpotential of different cancer cells in vivo (43). Thus, emerging
Fig. 2 Continued. B, analysis of HNSCC biomarkers in HNSCC-specificTMA. Head and neck ^ specificTMAs consisting of control andWD, MD, and PDHNSCC tumorsamples were stained for indicated proteins with appropriate antibodies as described in Materials and Methods. RepresentativeTMA cores are depicted (top). StainedTMAswere ‘‘scored’’ based on tissue differentiation and staining intensity. For cytokeratin 4 and cytokeratin16, light gray represents >5% and <25% of cells stained; mid gray,26% to 50% of the cells stained; darkgray, 51% to 75%; andblack,76% to100% of the cells stained. For vimentin and desmoplakin, the percentage of positive tumors for eachstage of differentiation (black box) compared with negative (white box) is depicted. In each case, the number of nonneoplastic (Normal) tissues analyzed was10, and thenumber of HNSCC cancer tissues is indicated.
The Oral Cancer Proteome
www.aacrjournals.org Clin Cancer Res 2008;14(4) February15, 20081013
evidence and our present findings suggest that this proteasemay play an unsuspected role in HNSCC progression.Taken together, the ability to combine laser capture
microdissection and in-depth proteomic analysis of formalin-fixed, paraffin-embedded tissues provides a wealth of informa-tion regarding the nature of the proteins expressed in normalsquamous epithelium and tumor progression. These proteinsinclude a number of tumor suppressor molecules and micro-RNA processing proteins likely involved in protecting theintegrity of the cellular genome in normal epithelial cells andtheir resident progenitor stem cells, as well as many moleculesinvolved in aberrant cell proliferation, survival, angiogenesis,proteolysis, and migration, whose contribution to tumorgrowth, resistance to treatment, and the metastatic spread ofHNSCC can now begin to be explored. The emerginginformation has also enabled identification of a large numberof proteins that are differentially expressed in normal oralsquamous epithelia and tumors exhibiting distinct differentia-
tion characteristics, thus representing suitable makers to studytumor progression. The future evaluation of the tumor makersdescribed in this study may afford an opportunity to exploretheir diagnostic and prognostic value, in particular for the earlydetection of HNSCC. On the other hand, we expect that ourstudy, documenting the successful use of proteomic techniquesand bioinformatic tools to analyze molecules expressed inarchival tumor and normal tissues, may now provide a proof ofprinciple that will boost ongoing systems to increase the scaleof data-generating proteomic efforts, which may ultimately leadto discovery of novel clinically relevant biomarkers andtherapeutic targets for HNSCC and other human malignancies.
Acknowledgments
We thank Thomas H. Bugge, Ana Raimondi, Panomwat Amornphimoltham,Chidchanok Leethanakul, RobertT. Dorsam,Thomas Guiel, and Marlene Darfler fortheir technical help, advice, and guidance.
References1. Jemal A, Siegel R,Ward E, et al. Cancer statistics,2007. CACancerJClin 2007;57:43^66.
2. Mao L, HongWK, Papadimitrakopoulou VA. Focuson head and neck cancer. Cancer Cell 2004;5:311^6.
3. ForastiereA, KochW,Trotti A, Sidransky D. Head andneck cancer. NEngl JMed 2001;345:1890^900.
5.Nagaraja GM, OthmanM, Fox BP, et al. Gene expres-sion signatures and biomarkers of noninvasive and in-vasive breast cancer cells: comprehensive profiles byrepresentational difference analysis, microarrays andproteomics. Oncogene 2006;25:2328^38.
6.Wulfkuhle JD, Liotta LA, Petricoin EF. Proteomicapplications for the early detection of cancer. Nat RevCancer 2003;3:267^75.
7. Canas B, Pineiro C, Calvo E, et al. Trends in samplepreparation for classical and second generation pro-teomics. JChromatogrA 2007;1152:235^58.
8. Fermin D, Allen BB, Blackwell TW, et al. Novel geneand gene model detection using a whole genomeopen reading frame analysis in proteomics. GenomeBiol 2006;7:R35.
9. Umar A, Luider TM, Foekens JA, Pasa-Tolic L.NanoLC-FT-ICR MS improves proteome coverageattainable for approximately 3000 laser-microdis-sected breast carcinoma cells. Proteomics 2007;7:323^9.
10. Baker H, PatelV, MolinoloAA, et al. Proteome-wideanalysis of head and neck squamous cell carcinomasusing laser-capture microdissection and tandem massspectrometry. Oral Oncol 2005;41:183^99.
11. AhramM, Flaig MJ, GillespieJW, et al. Evaluation ofethanol-fixed, paraffin-embedded tissues for proteo-mic applications. Proteomics 2003;3:413^21.
12.MolinoloAA, Hewitt SM, AmornphimolthamP, et al.Dissecting the Akt/mammalian target of rapamycinsignaling network: emerging results from the headand neck cancer tissue array initiative. Clin CancerRes 2007;13:4964^73.
13. Hood BL, Darfler MM, Guiel TG, et al. Proteomicanalysis of formalin-fixed prostate cancer tissue.Mol Cell Proteomics 2005;4:1741^53.
14. Ashburner M, Ball CA, Blake JA, et al.; The GeneOntology Consortium. Gene ontology: tool for theunification of biology. Nat Genet 2000;25:25^9.
15. LoWY,Tsai MH,Tsai Y, et al. Identification of over-expressed proteins in oral squamous cell carcinoma(OSCC) patients by clinical proteomic analysis.Clin Chim Acta 2007;376:101^7.
16. Kaur J, Srivastava A, Ralhan R. Expression of70-kDa heat shock protein in oral lesions: marker of
biological stress or pathogenicity. Oral Oncol 1998;34:496^501.
17. Li DQ, Hou YF, Wu J, et al. Gene expression pro-file analysis of an isogenic tumour metastasis modelreveals a functional role for oncogene AF1Q inbreast cancer metastasis. Eur J Cancer 2006;42:3274^86.
18. Sasaki T, Shiohama A, Minoshima S, Shimizu N.Identification of eight members of theArgonaute fam-ily in the human genome small star, filled. Genomics2003;82:323^30.
19. Leeman RJ, Lui VW, Grandis JR. STAT3 as a thera-peutic target in head and neck cancer. Expert OpinBiolTher 2006;6:231^41.
20. Amornphimoltham P, Patel V, Sodhi A, et al. Mam-malian target of rapamycin, a molecular target in squa-mous cell carcinomas of the head and neck. CancerRes 2005;65:9953^61.
21. Nakajima D, Okazaki N, Yamakawa H, et al. Con-struction of expression-ready cDNA clones for KIAAgenes: manual curation of 330 KIAA cDNA clones.DNARes 2002;9:99^106.
22. EliasJE, HaasW, Faherty BK, Gygi SP. Comparativeevaluation of mass spectrometry platforms used inlarge-scale proteomics investigations. Nat Methods2005;2:667^75.
23. ChandrashekarJ, Mueller KL, HoonMA, et al.T2Rsfunction as bitter taste receptors. Cell 2000;100:703^11.
24. Gaudin JC, Breuils L, HaertleT. Mouse orthologsof human olfactory-like receptors expressed in thetongue. Gene 2006;381:42^8.
25.Romio L,WrightV, Price K, et al. OFD1, the genemu-tated in oral-facial-digital syndrome type 1, isexpressed in the metanephros and in human embry-onic renal mesenchymal cells. J Am Soc Nephrol2003;14:680^9.
26. ChenYJ, Chang JT, Lee L, et al. DSG3 is overex-pressed in head neck cancer and is a potential molec-ular target for inhibition of oncogenesis. Oncogene2007;26:467^76.
27. Bartek J, Lukas J. Chk1and Chk2 kinases in check-point control and cancer. Cancer Cell 2003;3:421^9.
28. Powell SM, Zilz N, Beazer-Barclay Y, et al. APCmutations occur early during colorectal tumorigenesis.Nature1992;359:235^7.
29. Garnis C, Coe BP, Zhang L, et al. Overexpression ofLRP12, a gene contained within an 8q22 ampliconidentifiedby high-resolution array CGHanalysis of oralsquamous cell carcinomas. Oncogene 2004;23:2582^6.
30. Baek K, Brown RS, Birrane G, Ladias JA. Crystal
structure of human cyclin K, a positive regulator ofcyclin-dependent kinase 9. J Mol Biol 2007;366:563^73.
31.Wales MM, Biel MA, el DeiryW, et al. p53 activatesexpression of HIC-1, a new candidate tumour suppres-sor gene on17p13.3. Nat Med1995;1:570^7.
32. Rajalingam K,Wunder C, BrinkmannV, et al. Prohib-itin is required for Ras-induced Raf-MEK-ERK activa-tion and epithelial cell migration. Nat Cell Biol 2005;7:837^43.
33. Eldridge AG, LoktevAV, Hansen DV, et al. The evi5oncogene regulates cyclin accumulation by stabilizingthe anaphase-promoting complex inhibitor emi1. Cell2006;124:367^80.
34. Kalyankrishna S, GrandisJR. Epidermal growth fac-tor receptor biology in head and neck cancer. J ClinOncol 2006;24:2666^72.
35. Sehgal BU, DeBiase PJ, Matzno S, et al. Integrin h4
regulates migratory behavior of keratinocytes by de-termining laminin-332 organization. J Biol Chem2006;281:35487^98.
36. Basile JR, Afkhami T, Gutkind JS. Semaphorin 4D/plexin-B1 induces endothelial cell migration throughthe activation of PYK2, Src, and the phosphatidylino-sitol 3-kinase-Akt pathway. Mol Cell Biol 2005;25:6889^98.
37. Basile JR, Castilho RM,Williams VP, Gutkind JS.Semaphorin 4D provides a link between axon guid-ance processes and tumor-induced angiogenesis.Proc Natl Acad Sci US A 2006;103:9017^22.
38. Bertotti A, Comoglio PM,Trusolino L. h4 Integrin isa transforming molecule that unleashes Met tyrosinekinase tumorigenesis. Cancer Res 2005;65:10674^9.
39. Giordano S, Corso S, Conrotto P, et al. The sema-phorin 4D receptor controls invasive growth bycoupling with Met. Nat Cell Biol 2002;4:720^4.
40. Swiercz JM, Kuner R, Offermanns S. Plexin-B1/RhoGEF-mediated RhoA activation involves the re-ceptor tyrosine kinase ErbB-2. J Cell Biol 2004;165:869^80.
41. Reinheckel T, Hagemann S, Dollwet-Mack S,et al. The lysosomal cysteine protease cathepsin Lregulates keratinocyte proliferation by control ofgrowth factor recycling. J Cell Sci 2005;118:3387^95.
43. Liaudet-Coopman E, Beaujouin M, Derocq D, et al.Cathepsin D: newly discovered functions of a long-standing aspartic protease in cancer and apoptosis.Cancer Lett 2006;237:167^79.
Human Cancer Biology
www.aacrjournals.orgClin Cancer Res 2008;14(4) February15, 2008 1014
2008;14:1002-1014. Clin Cancer Res Vyomesh Patel, Brian L. Hood, Alfredo A. Molinolo, et al. ProgressionTissues: A Molecular Portrait of Head and Neck Cancer Proteomic Analysis of Laser-Captured Paraffin-Embedded